SBC-2 mandatory vs optional commands

Pat LaVarre p.lavarre at IEEE.org
Wed Jul 21 09:31:49 PDT 2004


* From the T10 Reflector (t10 at t10.org), posted by:
* Pat LaVarre <p.lavarre at ieee.org>
*
Kindly revised offline thank you:

>> Leave the 32 byte commands as strictly optional.  32 byte CDB support 
>> is
>> still very rare, as is support for XOR commands.
>
> Yes please.
>
> An
SBC
> that requires more than 12 bytes per CDB is an
SBC
> that won't fit into PATAPI/ SATAPI today.
>
> is my guess, because I've never seen an actual PATAPI allow more than 
> 12 bytes per CDB, even though since about the beginning the option for 
> as many as 16 was published.  And to allow more than 16, someone will 
> have to revise the PATAPI/ SATAPI texts.

Now continuing ...

> parallels "if any of
> WRITE (6)/(10)/(12) is implemented, WRITE (16) shall also be
> implemented." ...
>
> If T10 really wants to drive implementation of long LBA commands with
> SBC-2, then this is consistent; otherwise, we should revoke the
> mandatory state of READ (16) and WRITE (16).

I suppose a PATAPI/ SATAPI interpretation of this requirement will be 
to reject WRITE(16) as a CDB not copied out in full.  So yes, the 
device "implements" the WRITE(16) opcode, but sending that opcode 
doesn't work because the transport doesn't copy out the CDB in full.

Might be goodness to make that inevitable interpretation explicit in 
the t10.org text, before someone trips over it.  We could even suggest 
some specific SK ASC other than x 5 20 Unknown Op to indicate CDB not 
copied out the device in full.

> If T10 really wants to drive implementation of long LBA commands with
> SBC-2,

To what purpose?  Why require 64 or more bits/ LBA transport from a 
drive whose discs are all fully addressable with a 32-bit LBA?

Pat LaVarre

*
* For T10 Reflector information, send a message with
* 'info t10' (no quotes) in the message body to majordomo at t10.org




More information about the T10 mailing list